Skip to content

Commit 6311423

Browse files
SamTebbs33SamTebbs33
committed
Fix issue with reduction operand not being a phi
1 parent 124ecd3 commit 6311423

File tree

2 files changed

+167
-7
lines changed

2 files changed

+167
-7
lines changed

‎llvm/lib/Transforms/Vectorize/LoopVectorize.cpp‎

Lines changed: 6 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8803,8 +8803,13 @@ VPRecipeBuilder::getScaledReduction(PHINode *PHI,
88038803
return std::nullopt;
88048804

88058805
Value *Op = Update->getOperand(0);
8806-
if (Op == PHI)
8806+
Value *PhiOp = Update->getOperand(1);
8807+
if (Op == PHI) {
88078808
Op = Update->getOperand(1);
8809+
PhiOp = Update->getOperand(0);
8810+
}
8811+
if (PhiOp != PHI)
8812+
return std::nullopt;
88088813

88098814
auto *BinOp = dyn_cast<BinaryOperator>(Op);
88108815
if (!BinOp || !BinOp->hasOneUse())

‎llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll‎

Lines changed: 161 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1586,7 +1586,7 @@ for.exit: ; preds = %for.body
15861586

15871587
define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
15881588
; CHECK-INTERLEAVE1-LABEL: define i64 @dotp_cost_disagreement(
1589-
; CHECK-INTERLEAVE1-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0:[0-9]+]] {
1589+
; CHECK-INTERLEAVE1-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0]] {
15901590
; CHECK-INTERLEAVE1-NEXT: entry:
15911591
; CHECK-INTERLEAVE1-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
15921592
; CHECK-INTERLEAVE1-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 2
@@ -1617,10 +1617,10 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
16171617
; CHECK-INTERLEAVE1-NEXT: [[TMP15]] = add <vscale x 2 x i64> [[VEC_PHI]], [[TMP14]]
16181618
; CHECK-INTERLEAVE1-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
16191619
; CHECK-INTERLEAVE1-NEXT: [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
1620-
; CHECK-INTERLEAVE1-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
1620+
; CHECK-INTERLEAVE1-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
16211621
;
16221622
; CHECK-INTERLEAVED-LABEL: define i64 @dotp_cost_disagreement(
1623-
; CHECK-INTERLEAVED-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0:[0-9]+]] {
1623+
; CHECK-INTERLEAVED-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0]] {
16241624
; CHECK-INTERLEAVED-NEXT: entry:
16251625
; CHECK-INTERLEAVED-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
16261626
; CHECK-INTERLEAVED-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 4
@@ -1664,10 +1664,10 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
16641664
; CHECK-INTERLEAVED-NEXT: [[TMP25]] = add <vscale x 2 x i64> [[VEC_PHI1]], [[TMP23]]
16651665
; CHECK-INTERLEAVED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
16661666
; CHECK-INTERLEAVED-NEXT: [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
1667-
; CHECK-INTERLEAVED-NEXT: br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
1667+
; CHECK-INTERLEAVED-NEXT: br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
16681668
;
16691669
; CHECK-MAXBW-LABEL: define i64 @dotp_cost_disagreement(
1670-
; CHECK-MAXBW-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0:[0-9]+]] {
1670+
; CHECK-MAXBW-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0]] {
16711671
; CHECK-MAXBW-NEXT: entry:
16721672
; CHECK-MAXBW-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
16731673
; CHECK-MAXBW-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 8
@@ -1698,7 +1698,7 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
16981698
; CHECK-MAXBW-NEXT: [[PARTIAL_REDUCE]] = call <vscale x 1 x i64> @llvm.experimental.vector.partial.reduce.add.nxv1i64.nxv8i64(<vscale x 1 x i64> [[VEC_PHI]], <vscale x 8 x i64> [[TMP14]])
16991699
; CHECK-MAXBW-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
17001700
; CHECK-MAXBW-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
1701-
; CHECK-MAXBW-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
1701+
; CHECK-MAXBW-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
17021702
; CHECK-MAXBW: middle.block:
17031703
; CHECK-MAXBW-NEXT: [[TMP16:%.*]] = call i64 @llvm.vector.reduce.add.nxv1i64(<vscale x 1 x i64> [[PARTIAL_REDUCE]])
17041704
; CHECK-MAXBW-NEXT: [[CMP_N:%.*]] = icmp eq i64 16, [[N_VEC]]
@@ -1726,6 +1726,161 @@ exit: ; preds = %for.body
17261726
ret i64 %add
17271727
}
17281728

1729+
define void @not_dotp_not_phi2(ptr %matrix, i32 %n) #0 {
1730+
; CHECK-INTERLEAVE1-LABEL: define void @not_dotp_not_phi2(
1731+
; CHECK-INTERLEAVE1-SAME: ptr [[MATRIX:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
1732+
; CHECK-INTERLEAVE1-NEXT: entry:
1733+
; CHECK-INTERLEAVE1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[N]], 0
1734+
; CHECK-INTERLEAVE1-NEXT: br i1 [[CMP]], label [[FOR_PREHEADER:%.*]], label [[EXIT:%.*]]
1735+
; CHECK-INTERLEAVE1: for.preheader:
1736+
; CHECK-INTERLEAVE1-NEXT: [[LOAD_A:%.*]] = load i8, ptr null, align 1
1737+
; CHECK-INTERLEAVE1-NEXT: [[LOAD_A1:%.*]] = load i8, ptr inttoptr (i64 1 to ptr), align 1
1738+
; CHECK-INTERLEAVE1-NEXT: [[A_EXT:%.*]] = sext i8 [[LOAD_A]] to i32
1739+
; CHECK-INTERLEAVE1-NEXT: [[A_EXT1:%.*]] = sext i8 [[LOAD_A1]] to i32
1740+
; CHECK-INTERLEAVE1-NEXT: br label [[FOR_BODY:%.*]]
1741+
; CHECK-INTERLEAVE1: for.body:
1742+
; CHECK-INTERLEAVE1-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_PREHEADER]] ]
1743+
; CHECK-INTERLEAVE1-NEXT: [[PTR:%.*]] = phi ptr [ [[SCEVGEP:%.*]], [[FOR_BODY]] ], [ [[MATRIX]], [[FOR_PREHEADER]] ]
1744+
; CHECK-INTERLEAVE1-NEXT: [[ACCUM:%.*]] = phi i32 [ [[ADD_1:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_PREHEADER]] ]
1745+
; CHECK-INTERLEAVE1-NEXT: [[GEP_B:%.*]] = getelementptr i8, ptr [[PTR]], i64 1
1746+
; CHECK-INTERLEAVE1-NEXT: [[GEP_B1:%.*]] = getelementptr i8, ptr [[PTR]], i64 2
1747+
; CHECK-INTERLEAVE1-NEXT: [[LOAD_B:%.*]] = load i8, ptr [[GEP_B]], align 1
1748+
; CHECK-INTERLEAVE1-NEXT: [[B_EXT:%.*]] = sext i8 [[LOAD_B]] to i32
1749+
; CHECK-INTERLEAVE1-NEXT: [[MUL:%.*]] = mul nsw i32 [[A_EXT]], [[B_EXT]]
1750+
; CHECK-INTERLEAVE1-NEXT: [[ADD:%.*]] = add i32 [[MUL]], [[ACCUM]]
1751+
; CHECK-INTERLEAVE1-NEXT: [[LOAD_B1:%.*]] = load i8, ptr [[GEP_B1]], align 1
1752+
; CHECK-INTERLEAVE1-NEXT: [[B_EXT1:%.*]] = sext i8 [[LOAD_B1]] to i32
1753+
; CHECK-INTERLEAVE1-NEXT: [[MUL_1:%.*]] = mul nsw i32 [[A_EXT1]], [[B_EXT1]]
1754+
; CHECK-INTERLEAVE1-NEXT: [[ADD_1]] = add i32 [[MUL_1]], [[ADD]]
1755+
; CHECK-INTERLEAVE1-NEXT: [[SCEVGEP]] = getelementptr i8, ptr [[PTR]], i64 16
1756+
; CHECK-INTERLEAVE1-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
1757+
; CHECK-INTERLEAVE1-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[IV_NEXT]], [[N]]
1758+
; CHECK-INTERLEAVE1-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_EXIT:%.*]], label [[FOR_BODY]]
1759+
;
1760+
; CHECK-INTERLEAVED-LABEL: define void @not_dotp_not_phi2(
1761+
; CHECK-INTERLEAVED-SAME: ptr [[MATRIX:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
1762+
; CHECK-INTERLEAVED-NEXT: entry:
1763+
; CHECK-INTERLEAVED-NEXT: [[CMP:%.*]] = icmp sgt i32 [[N]], 0
1764+
; CHECK-INTERLEAVED-NEXT: br i1 [[CMP]], label [[FOR_PREHEADER:%.*]], label [[EXIT:%.*]]
1765+
; CHECK-INTERLEAVED: for.preheader:
1766+
; CHECK-INTERLEAVED-NEXT: [[LOAD_A:%.*]] = load i8, ptr null, align 1
1767+
; CHECK-INTERLEAVED-NEXT: [[LOAD_A1:%.*]] = load i8, ptr inttoptr (i64 1 to ptr), align 1
1768+
; CHECK-INTERLEAVED-NEXT: [[A_EXT:%.*]] = sext i8 [[LOAD_A]] to i32
1769+
; CHECK-INTERLEAVED-NEXT: [[A_EXT1:%.*]] = sext i8 [[LOAD_A1]] to i32
1770+
; CHECK-INTERLEAVED-NEXT: [[TMP0:%.*]] = zext i32 [[N]] to i64
1771+
; CHECK-INTERLEAVED-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 2
1772+
; CHECK-INTERLEAVED-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
1773+
; CHECK-INTERLEAVED: vector.ph:
1774+
; CHECK-INTERLEAVED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 2
1775+
; CHECK-INTERLEAVED-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
1776+
; CHECK-INTERLEAVED-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC]] to i32
1777+
; CHECK-INTERLEAVED-NEXT: [[TMP1:%.*]] = mul i64 [[N_VEC]], 16
1778+
; CHECK-INTERLEAVED-NEXT: [[IND_END1:%.*]] = getelementptr i8, ptr [[MATRIX]], i64 [[TMP1]]
1779+
; CHECK-INTERLEAVED-NEXT: br label [[VECTOR_BODY:%.*]]
1780+
; CHECK-INTERLEAVED: vector.body:
1781+
; CHECK-INTERLEAVED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
1782+
; CHECK-INTERLEAVED-NEXT: [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
1783+
; CHECK-INTERLEAVED-NEXT: [[VEC_PHI2:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[VECTOR_BODY]] ]
1784+
; CHECK-INTERLEAVED-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 16
1785+
; CHECK-INTERLEAVED-NEXT: [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 0
1786+
; CHECK-INTERLEAVED-NEXT: [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 16
1787+
; CHECK-INTERLEAVED-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[MATRIX]], i64 [[TMP2]]
1788+
; CHECK-INTERLEAVED-NEXT: [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[MATRIX]], i64 [[TMP3]]
1789+
; CHECK-INTERLEAVED-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i64 1
1790+
; CHECK-INTERLEAVED-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[NEXT_GEP3]], i64 1
1791+
; CHECK-INTERLEAVED-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i64 2
1792+
; CHECK-INTERLEAVED-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[NEXT_GEP3]], i64 2
1793+
; CHECK-INTERLEAVED-NEXT: [[TMP8:%.*]] = load i8, ptr [[TMP4]], align 1
1794+
; CHECK-INTERLEAVED-NEXT: [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
1795+
; CHECK-INTERLEAVED-NEXT: [[TMP10:%.*]] = sext i8 [[TMP8]] to i32
1796+
; CHECK-INTERLEAVED-NEXT: [[TMP11:%.*]] = sext i8 [[TMP9]] to i32
1797+
; CHECK-INTERLEAVED-NEXT: [[TMP12:%.*]] = mul nsw i32 [[A_EXT]], [[TMP10]]
1798+
; CHECK-INTERLEAVED-NEXT: [[TMP13:%.*]] = mul nsw i32 [[A_EXT]], [[TMP11]]
1799+
; CHECK-INTERLEAVED-NEXT: [[TMP14:%.*]] = add i32 [[TMP12]], [[VEC_PHI]]
1800+
; CHECK-INTERLEAVED-NEXT: [[TMP15:%.*]] = add i32 [[TMP13]], [[VEC_PHI2]]
1801+
; CHECK-INTERLEAVED-NEXT: [[TMP16:%.*]] = load i8, ptr [[TMP6]], align 1
1802+
; CHECK-INTERLEAVED-NEXT: [[TMP17:%.*]] = load i8, ptr [[TMP7]], align 1
1803+
; CHECK-INTERLEAVED-NEXT: [[TMP18:%.*]] = sext i8 [[TMP16]] to i32
1804+
; CHECK-INTERLEAVED-NEXT: [[TMP19:%.*]] = sext i8 [[TMP17]] to i32
1805+
; CHECK-INTERLEAVED-NEXT: [[TMP20:%.*]] = mul nsw i32 [[A_EXT1]], [[TMP18]]
1806+
; CHECK-INTERLEAVED-NEXT: [[TMP21:%.*]] = mul nsw i32 [[A_EXT1]], [[TMP19]]
1807+
; CHECK-INTERLEAVED-NEXT: [[TMP22]] = add i32 [[TMP20]], [[TMP14]]
1808+
; CHECK-INTERLEAVED-NEXT: [[TMP23]] = add i32 [[TMP21]], [[TMP15]]
1809+
; CHECK-INTERLEAVED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
1810+
; CHECK-INTERLEAVED-NEXT: [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
1811+
; CHECK-INTERLEAVED-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
1812+
;
1813+
; CHECK-MAXBW-LABEL: define void @not_dotp_not_phi2(
1814+
; CHECK-MAXBW-SAME: ptr [[MATRIX:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
1815+
; CHECK-MAXBW-NEXT: entry:
1816+
; CHECK-MAXBW-NEXT: [[CMP:%.*]] = icmp sgt i32 [[N]], 0
1817+
; CHECK-MAXBW-NEXT: br i1 [[CMP]], label [[FOR_PREHEADER:%.*]], label [[EXIT:%.*]]
1818+
; CHECK-MAXBW: for.preheader:
1819+
; CHECK-MAXBW-NEXT: [[LOAD_A:%.*]] = load i8, ptr null, align 1
1820+
; CHECK-MAXBW-NEXT: [[LOAD_A1:%.*]] = load i8, ptr inttoptr (i64 1 to ptr), align 1
1821+
; CHECK-MAXBW-NEXT: [[A_EXT:%.*]] = sext i8 [[LOAD_A]] to i32
1822+
; CHECK-MAXBW-NEXT: [[A_EXT1:%.*]] = sext i8 [[LOAD_A1]] to i32
1823+
; CHECK-MAXBW-NEXT: br label [[FOR_BODY:%.*]]
1824+
; CHECK-MAXBW: for.body:
1825+
; CHECK-MAXBW-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_PREHEADER]] ]
1826+
; CHECK-MAXBW-NEXT: [[PTR:%.*]] = phi ptr [ [[SCEVGEP:%.*]], [[FOR_BODY]] ], [ [[MATRIX]], [[FOR_PREHEADER]] ]
1827+
; CHECK-MAXBW-NEXT: [[ACCUM:%.*]] = phi i32 [ [[ADD_1:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_PREHEADER]] ]
1828+
; CHECK-MAXBW-NEXT: [[GEP_B:%.*]] = getelementptr i8, ptr [[PTR]], i64 1
1829+
; CHECK-MAXBW-NEXT: [[GEP_B1:%.*]] = getelementptr i8, ptr [[PTR]], i64 2
1830+
; CHECK-MAXBW-NEXT: [[LOAD_B:%.*]] = load i8, ptr [[GEP_B]], align 1
1831+
; CHECK-MAXBW-NEXT: [[B_EXT:%.*]] = sext i8 [[LOAD_B]] to i32
1832+
; CHECK-MAXBW-NEXT: [[MUL:%.*]] = mul nsw i32 [[A_EXT]], [[B_EXT]]
1833+
; CHECK-MAXBW-NEXT: [[ADD:%.*]] = add i32 [[MUL]], [[ACCUM]]
1834+
; CHECK-MAXBW-NEXT: [[LOAD_B1:%.*]] = load i8, ptr [[GEP_B1]], align 1
1835+
; CHECK-MAXBW-NEXT: [[B_EXT1:%.*]] = sext i8 [[LOAD_B1]] to i32
1836+
; CHECK-MAXBW-NEXT: [[MUL_1:%.*]] = mul nsw i32 [[A_EXT1]], [[B_EXT1]]
1837+
; CHECK-MAXBW-NEXT: [[ADD_1]] = add i32 [[MUL_1]], [[ADD]]
1838+
; CHECK-MAXBW-NEXT: [[SCEVGEP]] = getelementptr i8, ptr [[PTR]], i64 16
1839+
; CHECK-MAXBW-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
1840+
; CHECK-MAXBW-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[IV_NEXT]], [[N]]
1841+
; CHECK-MAXBW-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_EXIT:%.*]], label [[FOR_BODY]]
1842+
;
1843+
entry:
1844+
%cmp = icmp sgt i32 %n, 0
1845+
br i1 %cmp, label %for.preheader, label %exit
1846+
1847+
for.preheader: ; preds = %entry
1848+
%load.a = load i8, ptr inttoptr (i64 0 to ptr), align 1
1849+
%load.a1 = load i8, ptr inttoptr (i64 1 to ptr), align 1
1850+
%a.ext = sext i8 %load.a to i32
1851+
%a.ext1 = sext i8 %load.a1 to i32
1852+
br label %for.body
1853+
1854+
for.body: ; preds = %for.preheader, %for.body
1855+
%iv = phi i32 [ %iv.next, %for.body ], [ 0, %for.preheader ]
1856+
%ptr = phi ptr [ %scevgep, %for.body ], [ %matrix, %for.preheader ]
1857+
%accum = phi i32 [ %add.1, %for.body ], [ 0, %for.preheader ]
1858+
%gep.b = getelementptr i8, ptr %ptr, i64 1
1859+
%gep.b1 = getelementptr i8, ptr %ptr, i64 2
1860+
%load.b = load i8, ptr %gep.b, align 1
1861+
%b.ext = sext i8 %load.b to i32
1862+
%mul = mul nsw i32 %a.ext, %b.ext
1863+
%add = add i32 %mul, %accum
1864+
%load.b1 = load i8, ptr %gep.b1, align 1
1865+
%b.ext1 = sext i8 %load.b1 to i32
1866+
%mul.1 = mul nsw i32 %a.ext1, %b.ext1
1867+
%add.1 = add i32 %mul.1, %add
1868+
%scevgep = getelementptr i8, ptr %ptr, i64 16
1869+
%iv.next = add nuw nsw i32 %iv, 1
1870+
%exitcond.not = icmp eq i32 %iv.next, %n
1871+
br i1 %exitcond.not, label %for.exit, label %for.body
1872+
1873+
for.exit: ; preds = %for.body
1874+
%add.1.lcssa = phi i32 [ %add.1, %for.body ]
1875+
%add.float = sitofp i32 %add.1.lcssa to float
1876+
br label %exit
1877+
1878+
exit: ; preds = %for.exit, %entry
1879+
%result = phi float [ 0.000000e+00, %entry ], [ %add.float, %for.exit ]
1880+
store float %result, ptr %matrix, align 4
1881+
ret void
1882+
}
1883+
17291884
!7 = distinct !{!7, !8, !9, !10}
17301885
!8 = !{!"llvm.loop.mustprogress"}
17311886
!9 = !{!"llvm.loop.vectorize.predicate.enable", i1 true}

0 commit comments

Comments
 (0)